A key problem of modular neural networks is finding the optimal aggregation of the different subtasks (or modules) of the problem at hand. Functional networks provide a partial solution to this problem, since the inter-module topology is obtained from domain knowledge (functional relationships and symmetries). However, the learning process may be too restrictive in some situations, since the re...

A novel iterative approach based on a modular neural architecture [1] is presented for the classification of SAR images of sea ice. Additionally to the local image information the algorithm uses spatial context information derived from the first iteration of the algorithm and refines it in the subsequent iterations. The modular structure of the neural network is used with the aim to capture str...

Nearly all science and engineering fields use search algorithms, which automatically explore a search space to find high-performing solutions: chemists search through the space of molecules to discover new drugs; engineers search for stronger, cheaper, safer designs, scientists search for models that best explain data, etc. The goal of search algorithms has traditionally been to return the sing...

Assuring the security of networks is an increasingly challenging task. The number of online services and migration of traditional services like stocktrading and online payments to the Internet is still rising. On the other side, criminals are attracted by the values of business data, money transfers, etc. Therefore, safeguarding the network infrastructure is essential. As Intrusion Detection Sy...

This paper describes the application of a hierarchical modular neural network for an advanced approach solving the knock detection task for combustion engines. The knock detection is realized on a two-level feature extraction approach. It is not only based on the popular cycle-by-cycle classification but a tendency index for the knock condition is determined. The experimental environment consis...

Evolutionary algorithms are a promising approach for the automated design of artificial neural networks, but they require a compact and efficient genetic encoding scheme to represent repetitive and recurrent modules in networks. Here we introduce a problem-independent approach based on a human-readable descriptive encoding using a highlevel language. We show that this approach is useful in desi...

ÐIn this paper, we propose a new approach to combine multiple features in handwriting recognition based on two ideas: feature selection-based combination and class-dependent features. A nonparametric method is used for feature evaluation, and the first part of this paper is devoted to the evaluation of features in terms of their class separation and recognition capabilities. In the second part,...

It has been suggested that hippocampal rhythmical slow activity (theta rhythm) is related to cognitive process and the genesis of P300 response. To test this hypothesis, hippocampal EEG data from CA1 were recorded from rats trained to perform auditory discrimination oddball paradigm. In well-trained rats, signi"cant changes in the hippocampal theta rhythm were observed during an auditory oddbal...

This paper analyses two different approaches of fault distance location in a double circuit transmission lines, using artificial neural networks. The single and modular artificial neural networks were developed for determining the fault distance location under varying types of faults in both the circuits. The proposed method uses the voltages and currents signals available at only the local end...

A method of fault diagnosis using simple modular neural networks in a decision tree is proposed. The diagnostic accuracy of such a classifier is shown to be better than a single holistic neural network when applied to diagnosing faults in a seven component RC–network.